Where Fairchild had lumbered in the last few years, Noyce and Moore wanted Intel to race. “Use money to buy time because money is cheaper than time,” Noyce told Gene Flath, who had, of course, put this philosophy into practice when he bought equipment off the trade show floor. In the white-hot semiconductor industry of the late 1960s, being first to market would put Intel in the position of what Gordon Moore called “a rifleman who shoots at a blank wall, finds the bullet hole, and then paints the target around it.” The first company in the market always hits the bull’s eye because that company can draw the target around its own shot. Any late arrivals find the target already in place, the market defined by someone else.61
“ALL WE [HAVE] TO DO is reduce the cost [of semiconductor memories] by a factor of one hundred,” Noyce enjoyed reminding his employees in the first few months of Intel’s existence. The problem was not that a competitor was selling semiconductor memories far cheaper than Intel. The problem was that the existing storage technology for computers, an electromagnetic arrangement called “magnetic core memory,” sold for about 4 percent of what most people expected semiconductor memories would cost. Noyce wanted to get the cost of semiconductor memory down to about a penny per bit—even less than the cost of comparable core storage.62
Magnetic core memories resembled stacks of tiny square tennis racket heads strung with wire instead of cat gut. Iron donuts, each about the size of a pinhead, sat at every intersection of two pieces of wire. Currents sent through the wires could polarize the little doughnut in a particular direction—magnetized or not—to represent a binary one or zero. In other words, each donut could store one bit of information, which was transmitted to the computer through a series of transistors. Magnetic core memories were reliable, cheap, and perfectly adequate for the storage needs of late-1960s computers. Even with the relatively limited computer market of that era, core memories had netted hundreds of thousands of dollars in royalties for MIT, which held the patent.63
Magnetic cores had their shortcomings, however, and in these Noyce and Moore had seen a potential foothold for Intel. Cores were not a particularly fast means of storing data. The computer’s electronic pulses had to travel along about 10 square feet of wire to store 1,000 bits (zeroes or ones) of information. This traveling distance slowed processing, and as computers became more powerful and their storage needs grew, the wait to process and access data would become increasingly unacceptable. Moreover, the core memories were built by hand. Every one of those iron donuts was individually strung on a wire by a woman in a factory, most likely in Asia. Noyce and Moore knew that this labor-intensive means of production was not sustainable for a computer market growing exponentially, just as they had known a decade earlier that hand-wired discrete components could not serve the exploding market for space-age electronics.64
Moore and Noyce also knew that the problems with cores were irrelevant to most computer engineers, who did not spend time thinking about how they would build their machines ten years in the future. These engineers cared that their computers work now, and so the cost advantages of semiconductor memory would have to be overwhelming before the engineers would consider abandoning the clunky, but reliable, magnetic cores. A sense of déjà vu may have again struck Noyce and Moore, who faced a similar obstacle when they initially brought the integrated circuit to market. Noyce, the architect of Fairchild’s decision to sell integrated circuits below cost to get a foothold in the discrete components market, was betting a similar strategy would work for semiconductor memories. Hence the mantra: “all we have to do is reduce the cost.”
But, of course, that was not all they had to do. First they had to build the memories. A group consisting of Noyce, Moore, Grove, Hoff, Graham, and occasionally two or three others, had begun developing formal product plans in August 1968. They decided to take a three-pronged approach. Dick Bohn, a mercurial, brilliant, intense researcher from Sylvania who held several key patents in the bipolar field, would lead the effort to build a bipolar memory device. This would be a very fast chip that would drive other memory devices in a computer. Bipolar techniques had been around since the earliest days of Fairchild, and building them was as close to rote as anything in the semiconductor industry. But no one had squeezed as many transistors on a bipolar circuit as Intel proposed to do.65
Andy Grove and Les Vadasz would open up a second line of attack by trying to build a memory built with a much newer technique called MOS (metal-oxide-silicon). It takes fewer steps to build MOS integrated circuits than bipolar circuits, and MOS devices could be much more densely packed with transistors, which made them the logical choice for complex chips. That, at least, was the theory. In reality, MOS devices were slower, more temperamental, and even more dependent on “black magic” than other semiconductor devices. MOS circuits had yet to be produced in volume and existed primarily as lab curiosities. Here Intel planned to start slow, building a small memory targeted to serve functions too simple for cores.66
A third group attempted to assemble four small memory chips into a single module that would function as a large-but-cheap memory. This memory was called the “multichip” or sometimes the “flip chip” because it was turned over, front to back, in the assembly process. The flip chip intrigued Noyce, not the least because it offered the possibility of eliminating one bonding step from the process of building memory chips. He focused his lab efforts on the die attacher that picked up the chip, placed it into a package, and slightly pressed down on the chip to attach it to the package. It was not easy to build a machine that did not crush or crack the chip at some point during this process.67
Noyce and Moore developed the three-pronged attack on the assumption that if one approach did not work out, maybe another one would. Moreover, the bipolar and MOS approaches could share many techniques and facilities, which meant that the company’s effort, when it came to buying equipment and developing operating procedures for growing crystals, cutting wafers, maintaining cleanliness, and the like, was not as fractured as might appear on first blush. If, in the best-case scenario, more than one process yielded success, each was targeted to serve a different market and to work in tandem with the other processes.68
Noyce spent more time in the lab during the first year at Intel than he had in the previous eight years at Fairchild, yet he offered no insights that approached the level of his integrated circuit ideas and no far-out suggestions that worked in the way he anticipated—though he did have one idea, for using blown diodes to make a type of memory device, that eventually led Ted Hoff to a related patent. Noyce’s work on the flip-chip bonder was more a shop-class than physics-lab type of job. As he had feared, Noyce had indeed been gone from the lab too long to make a meaningful technical contribution.69
But his technical instincts were still excellent. “His questions were so perceptive,” recalls Ted Hoff. “No matter how deep you went, he was right there, always caught up and then racing ahead. It is kind of a shame he was so successful as a businessman that it kept him away from the technical side.” Says Les Vadasz, “He challenged you at the time. Why can’t we do this? How come we do this? What about this? … He really pushed you to push yourself. If you were talking about a 512-bit memory, [Noyce would ask] why not 1K? If you’d say this speed, he’d say, why not this speed? He just pushed and pushed and pushed. Even though he may not have known the details about the area he was pushing, he had enough conceptual understanding that he pushed you to the point where you had to question your own basis for why you’re coming up with a conclusion. I think that had an impact on all of us.”70
To be “pushed” by Noyce in this way was not an altogether pleasant experience. It was a bit like being peeled, coolly and efficiently stripped down, layer by layer, until Noyce reached what he considered the heart of the matter. The person whom Noyce “pushed” would leave the conversation strangely invigorated by a sense that he had sloughed off all superfluous accretions of conservative thinking and conventional wisdom—and that he could now do wh
at Noyce had somehow convinced him it was possible to do.
After observing William Shockley’s methods of using simplifying assumptions to speed up his company’s research, Noyce had come to believe that scientists could approach their work in two very different ways. Researchers could adopt the “pretty” approach, in which they devote a great deal of time and effort to developing a technique or machine that will allow them to test their ideas with exact measurements that yield final definitive answers. Or a researcher could try the “quick and dirty” way, moving forward with an idea as soon as a rather rudimentary test indicates it will probably work. Noyce believed that the quick-and-dirty method generated “90 percent of the answer in 10 percent of the time.” He disdained the pretty method as “a bit like telling a soccer player never to kick the ball until you have an ideal shot all carefully lined up and know exactly how hard to kick the ball. Ninety minutes may be over before you locate that opportunity.”71
The quick-and-dirty research method, which Moore nicknamed “the Noyce principle of minimum information,” prevailed at Intel, which was, like its founders, forever in a hurry. Even at Intel speed, it took five or six man-years to design a new circuit. And there was another benefit to operating on the Noyce principle, according to Moore. The company produced few spin-offs “because it does not generate a lot more ideas than it can use.”72
Andy Grove was the ideal man to implement the Noyce principle of minimum information. Noyce, who called him “the whip,” once told a friend that one of the many reasons he was happy to have Grove at Intel was that “it is tough for me to do the hard things, but it’s not tough for Andy.” To succeed with Grove, an early employee recalled, one needed, above all, “to do something…. No soul-searching, no thumb twiddling. [Grove was] keeping it moving, Hannibal through the Alps, keep the elephants moving, don’t let them go down on their knees.”73
Grove quickly moved from directing the lab to overseeing all aspects of research and manufacturing as manager of operations. He tracked everything, trusted nothing and no one, yelled and cursed to get his point across, frustrated others to the point that they yelled and cursed to get their points across, and insisted that people state—often in writing—exactly what they were going to do and when. And always, always, Grove followed up to confirm people kept their commitments. In later years, he would require all employees in his area of responsibility who arrived at work after 8 AM to sign a “Late List,” which was then circulated to managers and supervisors. He once called a plan to test Intel’s finished, ready-to-ship product with the rigorous high-temperature, high-voltage procedures that most companies used only for the first batches of chips “the best thing since the Pill.”74
He could not have been more different from Noyce. Their divergent approaches caused more irritation for Grove than for Noyce. Grove, a linear thinker, found it hard to follow Noyce’s thinking, which reminded him of “a butterfly hopping from thought to thought.” Occasionally these ideas were intriguing enough to distract Grove from the work he had set for himself, but even more maddeningly, they would distract his employees, whom Grove would occasionally find had launched themselves on a tangent to chase down one of Noyce’s more tantalizing mind flashes. For his part, Noyce had little problems with the differences between his and Grove’s approaches to business. Noyce believed that for a company to be successful, it “must keep the vision of product and direction very narrow, while keeping the peripheral vision of market forces affecting the business very broad.” Noyce would provide the necessary wide ranging vision and Grove would provide the equally necessary precision focus.
In truth, Grove and Noyce had little to do with each other in the earliest days of Intel. Noyce was outside of the office as much as he was in it, talking to the press, potential customers and investors, and job candidates. One executive hired in January 1970 says, “I don’t think I talked to, or even met, Bob for months. He was always just whipping through.” Most of Noyce’s direct reports soon found surrogate bosses to whom they could turn for daily feedback. “I guess I officially reported to Bob,” mused one such employee, “but I really reported to Andy and Gordon.” Explained another, “Nobody actually reported to Bob. Nominally some people did report to both Bob and Gordon, but Gordon was always on the scene, and Bob was never there.”75
With Noyce turning his attention outward, Grove focused on the inner workings of Intel, the essential minutiae of day-to-day operations. Moore bridged the gap between the two, weighing Noyce’s signals from the outside against research in the field and Grove’s reports from the inside. One person familiar with the trio described their roles this way: “Noyce would say, ‘Some day we will use semiconductors to perform [some outrageous job].’ Then Moore would say, ‘To do that, we would need to transcend technical problems Y and Z.’ And Grove would say, ‘That means we’ll need to get however-many-more engineers and increase yields by X percent and plant space by Y feet.’ Their strengths just balanced each other so well.” Their roles, in many peoples’ opinions, could be described even more succinctly: Mr. Outside (Noyce), Mr. Inside (Moore), and Mr. Implementation (Grove).76
Intel meant something different for Grove than it did for Moore or for Noyce, who once said he started Intel, in part, to see “whether [we] could do it again; whether [Fairchild’s success] was just plain dumb luck the first time around.” Noyce and Moore had already proven their worth at Fairchild. Intel represented Grove’s chance to do the same. Grove has said that the months he spent watching Fairchild R&D’s “outstanding work” languish in the purgatory between development and manufacturing “fill[ed] me with self doubts to the point that I was wondering if the work we were doing was as good as I thought it was.” In his last year at Fairchild, Grove had been convinced that the Simon and Garfunkel hit song “Faking It” was directed at him.77
Grove and marketing manager Bob Graham were at constant loggerheads. Grove thought Graham was trying to do his job—every day, Graham went down to the room where the designers worked to “check on the progress of the product”—and besides, Grove had “almost no use” for marketing. Grove figured that if Intel built an excellent product (his responsibility, thank you very much) the customers would not need too much convincing to buy it.78
For his part, Graham thought Grove, a research scientist until a few months before, did not know enough about manufacturing to be running operations. It further incensed Graham that Grove paid no attention to his explanations of customers’ expectations for Intel products. Graham also thought that for someone who had “no use” for marketing, Grove spent a great deal of time nitpicking the wording on Graham’s proposed advertisements and press releases.79
The real problem between Graham and Grove was far more fundamental than their fine-pointed accusations, however. Only one of these two men could be the second-in-command to the founders. Too much was at stake for both Graham and Grove for them to tolerate each other’s presence in the company for long. Within months of Intel’s founding, the enmity between these top two lieutenants threatened to poison relationships throughout the company.80
8
Takeoff
The first three years that Noyce spent at Intel were among the most exhilarating of his career. He enjoyed his return to the jack-of-all-trades he had been in the early days of Fairchild. He regularly sat in on technical reviews, led the weekly executive staff meeting and a product planning session every two weeks, helped a technician build a tool she thought would make her work easier, met with promising recruits, mopped the floor when a pipe broke, and delivered a number of lectures at technical and business conferences on topics such as “What Happened to LSI?” He and Moore set aside one lunch hour each week to eat with small groups of employees, and a morning every month to talk to industry analysts.
Noyce’s interactions with potential customers in the first months of Intel’s existence revolved around determining market needs. He spoke with senior managers at Burroughs, Honeywell, and Univac, firms interested in semi
conductor memories for the computers they were building, and at Memorex, whose executive vice president, Jim Guzy, had recently joined the Intel board. Occasionally Noyce would explain what Intel was trying to build and ask for suggestions for improving their technical plan. That Noyce believed the need for feedback overrode the risk that a customer might share Intel’s plans with competitors is perfectly consistent with his instinctive trust in people and his determination to build products that people would actually buy.1
The steady stream of phone calls Noyce had fielded from his study at home became a constant rush of visitors to his Mountain View office. As Grove recalls it, “nobody in the industry would move from a job, or start a company, or do much of anything, without paying homage to Bob and getting his blessing. On top of this, there were customers, bankers, directors, people connected with the financing…. It was funny. All these people in suits, coming and going. Noyce was the ‘go-to person’ of the day.”2
Even his family picked up on his excitement. Noyce would at times bring the children with him to the office on weekends. He put license plates that read “INTEL” on his new car, a white Cougar. One memorable day, the entire family helped to cut rubylith, the red cellophane-like sheets that operators laid over a paper design of a circuit and cut according to the paper pattern. (The rubylith would then be shrunk to the proper size and used to make the mask for the circuit.) The children—now young adolescents—enjoyed a fine few hours wielding exacto knives and “helping Da,” who was also hard at work over the light table, along with Betty. At dinner, Noyce would often talk about Intel’s struggles and successes, and the children felt enough a part of the company that Penny burst into a neighbor’s house shouting “They made one, they made one!” after her father told her that Intel’s first memory chip had been successfully tested.3
The Man Behind the Microchip Page 27
